Week 10: Hands-on Deep Learning II
Lab
Overview
The second hands-on lab session on deep learning. In Week 9 the network predicted a number (regression). This week it answers a yes or no question: binary classification. You build a multilayer perceptron that classifies breast tumors as malignant or benign, and three things change compared with the earlier labs:
- the output of the network is turned into a probability (logit and sigmoid),
- the loss function becomes binary cross-entropy,
- the evaluation uses the confusion matrix, precision, and recall, and you tune the decision threshold to trade false alarms against missed cancers.
How to run the lab in Google Colab
- Download the notebook and the dataset from the table below.
- Open colab.research.google.com, choose File, Upload notebook, and select the downloaded notebook.
- Choose Runtime, Run all. When the upload dialog appears, select the dataset file
breast_cancer.csvfrom your computer. - Read the text between the cells: it explains what each step does and why. Finish with the exercises at the end of the notebook.
No GPU is needed; everything runs on the CPU in under a minute.
Materials
| Material | Where |
|---|---|
| Lab 3 notebook | lab03_classification_mlp.ipynb |
| Dataset | breast_cancer.csv |
Dataset citation: Wolberg, W., Mangasarian, O., Street, N., and Street, W. (1993). Breast Cancer Wisconsin (Diagnostic) [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5DW2B.